One wonders how much cheaper, allowing for size of production run, a guided, gun-launched projectile is than a pocket-propelled one with the same level of performance.
Similar to or more expensive per kill.
Depending on brains in the shell, the quick references I found for M982 Excalibur and Ground Launched GBU39 put the advantage well in favor of the cannon projectile, even with recycling existing and paid for rocket motors for the GLSDB.
You found some weird references, because GBU-39 is cheaper than M982, as of FY2019. $40,000 USD versus $85,000 USD.
A Stormbreaker costs twice as much as M982 but can hit moving targets without external guidance. It's still cheaper than a Javelin, because it's a glide bomb deployed by aircraft. Excalibur is a single 155mm shell with an INS kit. It's $70,000. Copperhead was similarly expensive, but can hit moving targets, but also requires a ground designator team which needs to be emplaced and in the proper position...
A self-guiding Copperhead would probably cost similar to Stormbreaker or Javelin, so six digits (maybe twice as much as Excalibur), because it requires shock hardened electronics.
Once you factor in the ancillary equipment like laser designators, or amount of munitions needed to be expended to hit mobile targets, the advantage of the single unit cheaper shell tends to evaporate. Per pound of explosive, a missile or rocket will always deliver more killing power for less cost, at least when it comes to delivering ordnance against wide area targets, than shells.
This is one of those immutable things and is the main reason why multiple rocket launchers have survived for 75+ years in military use.
Rockets are better for delivering large blast-type warheads to kill particularly annoying things, like warships or reinforced structures, while cannons and aviation bombs are better at delivering penetrating warheads due to the high speeds they impact. It would be hard to fit a reinforced bunker buster into a GMLRS but trivial to fit a blast-frag warhead or DPICM, which can annihilate soft targets like C3I and SAM sites.
GROUND LAUNCHED SDB, not a basic GBU-39.
How do you have figures about something that doesn't exist? No one has purchased GLSDB, therefore it has no recorded cost...
Comparing the non-existent GLSDB and the M982 is disingenuous. One doesn't exist and the other has a third the range (150 vs 50 km), so it's apples to oranges in two ways. GLSDB would have a range comparable to the Navy's aborted LRLAP, which costs around $1 million FY2016 which are the latest figures, but I think GLSDB would have cost less than the near "Tomahawk missile" costs of the 100 nmi ranged CLGP.
Anyway, this thread is about Starstreak. If you have to pick between making laser guided 25mm shells or slapping Starstreaks on the side of a vehicle, you'll choose the Starstreak if you're sensible and of sound mind that is motivated by combat performance. People have different motivations, and different competencies, so they choose different things. For the most part, CLGPs are compromised in unit cost and engagement cost compared to maneuvering missiles.
Starstreak's darts are around 20-25mm diameter and the missile is from the 1980's. However, there's a reason no one was rushing to get 35mm, subcaliber, laser guided HTK shells into a Gepard when they could just put Stingers on the sides at the time. It isn't because Germany couldn't do this. It's because CLGPs are typically the worse option.
The U.S. military vacillates between generally cogent decisions (Excalibur, Copperhead, PGK) that fit the peculiarities of the CLGP to decisions that make people question the motivations of the acquisitions officers (ALaMO, LRLAP, SLRC).
There seems to be the idea that CLGPs are cheaper than rockets, which sort of meanders through the E-Ring every few years, and this might be because people are comparing things like TOW and Copperhead or Excalibur with Javelin, or someone is suggesting a 50 gram explosive warhead is comparable to a 5,000 gram warhead against small craft, or whatever. This filters through to the press and makes ordinary laymen think the same. It's a bit bizarre. CLGPs typically require multiple engagements per attack, not because they don't hit, but because their explosive masses are so much smaller than a rocket.
Rockets carry bigger warheads which means more killing. Shells carry smaller warheads which means less killing.
Thinking this way tends to result in the belief of the superiority of the MRL to the cannon writ large, which is probably true, but there aren't enough MRLs in anyone's inventory to fully displace the cannon and the minimum range issues remain. In a world with infinite resources and instant timescales, I guess large caliber MRLs (220-300mm) and 120mm mortars would be the only pieces of the artillery park since they maximize the ability to kill the enemy.
Obviously you work with what you have, though. CLGPs are a good way to give precision guidance to a howitzer, one that would otherwise require two or three minutes to fire a dozen shells when it can fire two shells at the same target in 30 seconds for the same price and similar effectiveness instead.
They aren't good for engaging small targets in general though. They tend to cost as much per engagement as dumb rounds, you're just taking less time, expressed by fewer shells, to kill the target. You're still paying the same amount of money.
Nuclear weapons are the only time this shifts in favor of the attacker, I think. Even a small thermonuclear weapon can destroy much more value for money than the $2-4 million it nominally costs.